Fintel Study Finds Affiliates Shape AI Financial-Product Recommendations

Credit-union marketing teams now have a concrete AI-search measurement problem: find out whether assistants can identify the institution's products, which pages they cite, and whether the resulting descriptions remain accurate and appropriately disclosed. A new financial-marketing study suggests that affiliate pages can have outsized influence on the answers consumers receive.

Fintel Connect said October 7 that its study examined 2,400 responses and 7,809 citations across ChatGPT, Claude and Gemini. The prompts covered consumer and business credit cards, personal loans and high-yield savings accounts.

The company reports that 90% of decision-stage prompts produced a citation, compared with 8.8% of awareness-stage prompts. It also says affiliate pages accounted for more than 54% of citations in recommendation answers, while half of all responses in the sample cited no source.

Those figures come from a company that operates an affiliate-marketing platform for financial institutions. CreditUnionAI News did not independently reproduce the study, and the public release does not provide enough detail to assess every prompt, model version, sampling date or category-level denominator. The findings are a useful discovery signal, not an industry benchmark or proof that affiliate placement causes a brand to appear.

Measure the answer before buying the channel

A credit union can build a bounded audit without sending member data into a public model. Start with a fixed set of non-personal prompts that reflect real product-research tasks: local savings options, auto loans for a defined credit profile, small-business cards, or comparisons of fees and eligibility. Record the model, date, prompt, brands named, cited URLs, quoted terms and any missing qualification.

Run the same set on a schedule and score four separate outcomes: whether the credit union appears, whether a source is cited, whether the product facts are accurate and current, and whether the result sends a qualified visitor to an owned page. Visibility alone is not success. An assistant can name a product while misstating membership eligibility, APY conditions, fees, geography or availability.

The institution should also separate owned, earned and paid influence. A citation to the credit union's product page, an independent comparison article and a compensated affiliate placement have different governance, disclosure and measurement requirements. The FTC's endorsement guidance is a useful baseline for reviewing material connections and clear disclosures, though legal and compliance teams should determine what applies to a particular arrangement.

Treat affiliates as governed distribution partners

If an affiliate participates in the credit union's acquisition funnel, marketing and compliance leaders need an inventory of approved publishers, claims, product feeds, tracking links and disclosure requirements. Establish who updates rate and fee information, how quickly outdated pages must be corrected, which keywords or comparisons are prohibited, and how the institution can suspend traffic when a page becomes misleading.

The same controls should reach analytics. Attribute sessions and completed applications to the source that actually referred them, preserve campaign and landing-page identifiers, and reconcile approved affiliate records with observed traffic. Do not infer causation from a brand mention, and do not let a vendor's platform totals substitute for the credit union's own qualified-session, application and funded-account evidence.

Credit unions should also keep member-level data out of open-ended prompt testing. If a later program connects AI-search exposure to personalization, the institution needs a documented consent basis, purpose limitation and suppression path. Our coverage of category-specific communication consent shows why a member's choice must follow the message and channel rather than disappear inside a general growth score.

Build evidence that survives model churn

AI-search results can change when models, retrieval systems, ranking rules or source indexes change. That makes durable content hygiene more valuable than one favorable screenshot. Keep product facts in accessible, crawlable pages; state eligibility and material conditions clearly; date rate information; maintain canonical URLs; and remove or redirect expired offers.

Then review the answer set for member impact. If the assistant disproportionately recommends products with affiliate economics, the credit union should not imitate that incentive blindly. Compare acquisition cost, approved and funded-account rates, early account activity, complaints, opt-outs and attrition. The Dort Financial personalization case illustrates the broader operating requirement: connect targeting and offers to measured member outcomes, not just message volume.

Fintel's study makes AI-search visibility measurable enough to test, but not mature enough to manage by impression alone. The defensible credit-union response is a source-level audit that ties citations to accurate product information, governed partners and qualified member outcomes.